Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

X-ray computed tomography for in-situ inspection of CMC material under load

Eleri Williams, Alan Clarke and P. Ian Nicholson

TWI Technology Centre Wales, Harbourside Business Park, Harbourside Road, Port Talbot, SA13 1SB

James Kern and Georgios Asfis
TWI Ltd, Granta Park, Great Abington, Cambridge

Martin Bache
Institute of Structural Materials, Swansea University, Swansea, SA1 8EN

Insight (Journal of BINDT)


An understanding of damage evolution within ceramic matrix composites (CMCs) is important if they are to be utilised to their full potential within industrial applications. To this end, the use of X-ray computed tomography (XCT) has been shown to be useful in detecting sub-critical damage and ultimate failure mechanisms in a range of CMCs. The observation of inherent processing artefacts and the tracking of any resultant damage in the structure of CMCs is an important step toward understanding the behaviour of these materials. Previous work has seen mechanical loading of a sample applied in a separate test rig and then removal of the sample for subsequent inspection using XCT. However, due to the complexity of the failure mechanisms in CMCs, a comprehensive damage analysis is still a challenging problem. An important step toward a better understanding of the behaviour of these materials is to perform in-situ XCT inspection, i.e. whilst the material remains under load. This paper considers the requirements for developing XCT for in-situ inspection of CMC material. The motivation was to perform volumetric inspection of mechanically loaded CMC samples using XCT and capture the evolution of damage within the material under increasing levels of monotonic tensile stress. An initial XCT investigation of a CMC sample, using a commercially available in-situ mechanical loading rig, is presented.

1. Introduction

Ceramic matrix composites (CMCs) are typically composed of ceramic fibres embedded within a ceramic matrix. When compared to conventional, monolithic ceramics, CMCs will display an increased degree of toughness. Early stage cracks predominantly form in the matrix of the material but do not propagate into the fibres, instead the fibres bridge the cracks, supporting the remote load up to the point where individual fibres or bundles of fibres fail during the onset of ultimate failure(1). Variation of the fibre to matrix volume ratio, lay-up configuration, fibre composition and in particular the fibre/matrix interface/coating can control the constitutive and fracture properties of the CMC.

The application of CMCs in engineering components relies on the understanding of damage accumulation under various loading configurations. The behaviour of CMCs can be affected by the reinforcement architecture and associated artefacts introduced during the manufacturing process. In order to characterise these features within the body of the material, and subsequently track damage evolution under loading, non-destructive evaluation (NDE) has been utilised. There are many types of NDE available and many of these have been applied to CMC inspection. These include ultrasonic testing(2), acoustic emission testing(2), thermography(3) and dye penetrant testing(4). In particular x-ray computed tomography (XCT) has shown to be effective in improving the understanding of damage and failure mechanisms of CMCs as well as their evolution when under load. The technique is also capable of volumetric inspection of the part being tested.

The ability to characterise the material in three dimensions as opposed to two in traditional radiography, is a distinct benefit. Johnston et al(5) observed that 2D radiography was not capable of identifying planar defects with no observable change in thickness, whereas 3D XCT scanning was able to detect the same features and de-laminations within a CMC material. Having a 3D image of damage evolution can be useful in the validation of computer simulations. Saucedo-Mora et al(6) used CT data to validate a novel finite element microstructural adaptive mesh-free model that allows the microstructure of CMCs to be introduced into a continuum finite element model in order to simulate the fracture mechanics of CMCs more accurately.

In this context, this paper investigates the damage evolution in a CMC mechanical test sample using in-situ XCT, i.e. whilst the sample remains under load, using a commercially available in-situ tensile tester. The sample was subjected to progressively increasing levels of monotonic tensile stress. XCT images were taken after each loading step to identify any new damage and to track any previous areas of interest. This advances on previous work carried out by the collaborating partners where XCT has been applied as a “post mortem”, post mechanical testing and unloading of the test specimen(7).

2. Experimental Methods

2.1 XCT instrumentation

The XCT system used to obtain XCT scans during this study was an HMX 225 originally supplied by Metrix X-Tek Ltd. The system comprised of:

  • A microfocus X-ray source based on cone beam geometry with four different anode targets available (tungsten, silver, molybdenum and copper). Minimum focal spot size of 5-10 µm, a maximum tube potential of 225 kV, and maximum tube current 2mA. Various filtering materials can be applied.
  • A 5-axis rotating manipulator capable of handling up to 25kg load and geometric magnification up to 160x.
  • A detector system consisting high resolution 16-bit flat panel digital X-ray detector with 2000x2000 pixels and 200 µm pixel size.
  • The source, manipulator, detector and XCT data acquisition were controlled by X-Tek InspectX software, which digitised the radiographic images to 16-bit format.
  • The XCT reconstruction software was provided by Metris X-Tek (CTPro). The XCT visualisation software was VGStudioMax, generating CT images in 16-bit format.

2.2 CMC sample preparation

The CMC specimen was extracted from a flat plate, approximate thickness 5mm, comprising a 12 ply 0-90˚ woven architecture of reinforcing fibre bundles.  A central, parallel sided gauge section of 5mm x 10mm was machined. Tab ends allowed an interface to a pinned keep plate grip design to either end. This nominal CMC material was selected simply as a means towards the assessment of the experimental technique under consideration and the general details of deformation mechanisms.

2.3 Specimen loading and XCT setup

The tensile tests were performed at room temperature, with the CMC specimen loaded using a commercially available tensile and compression electro-mechanical testing machine. The rig, available from Deben is the CT5000 system. It is capable of loading up to 5000N and tracks the applied stress and strain of the part via coordinated software. Strain can be calculated as a function of actuator displacement divided by the original length of the parallel gauge. The Deben frame was placed on the XCT turn table inside the X-ray booth as shown in Figure 1. The loading force of the rig is transferred into the specimen under test via a cylindrical, glassy carbon tube. The tube material offers low-attenuation to X-rays and the tube geometry allows for a full 360˚ “line of sight” for X-ray inspection of the specimen whilst under load.  In order to acquire the CT data, 2000 2D X-ray images are acquired during the 360˚ rotation of the rig and specimen.

[ Zoom ]
Figure 1. (a) The Deben loading rig within the X-ray booth with the detector in view and (b) relative to the X-ray source.
Figure 1. (a) The Deben loading rig within the X-ray booth with the detector in view and (b) relative to the X-ray source

The CMC specimen was inspected by XCT post machining but prior to test as a reference. A monotonic load ramp was applied during the experiment at a nominal rate of 1kN per minute. Loading was interrupted after four successive load steps to allow XCT scanning, i.e. at 25% of the applied Pmax, 50%Pmax, 75%Pmax and Pmax. During these hold periods, approximately 20 minutes each in duration, the Deben rig transferred to displacement control to keep the specimen stationary during x-ray acquisition. The X-Ray settings used for each scan can be seen in Table 1.

Table 1 . XCT parameters used for scanning the CMC specimen

X-Ray Parameter




Current (µA)



1mm Aluminium

Exposure time (ms)




Source to object distance (mm)


Source to detector distance (mm)


Voxel size (mm)


3. CT detection of damage accumulation

As a general reference of the CMC architecture, a three dimensional XCT image is presented in Figure 2. This has been reconstructed from the acquired XCT volumetric data acquired for the central gauge section of the specimen and has been recorded prior to the loading experiment. The orientation for the front view plane is highlighted and it is the image slice views in this plane that are analysed in this paper. The network of 0-90˚ reinforcing fibre bundles can be clearly distinguished, although at this magnification it is not possible to resolve individual fibres. The critically stressed region of the specimen is seen to sample approximately five longitudinal and seven transverse fibre bundles within the parallel gauge. The as processed surface finish from the original CMC plate provides an irregular topography to the front (and rear) faces whilst the lateral faces display a relatively smooth machined finish.

[ Zoom ]
Figure 2. Three dimensional view of the specimen gauge section prior to loading
Figure 2. Three dimensional view of the specimen gauge section prior to loading

XCT image slices selected for the same layer in the XCT volume data are then presented in Figure 3 taken during the four load steps applied up to Pmax. The plane of view is approximately 1mm below the arbitrary “front” face in each case. Comparisons between the fibre architecture and persistence of the early stage features give confidence that the same plane has been selected throughout the sequence. A network of predominantly transverse cracking is noted, with the number of individual cracks increasing and length of any pre-existing cracks lengthening with increasing load. Crack opening is widest for those cracks breaking the lateral edges of the specimen. At the point of maximum loading, some individual cracks fully traverse the width of the gauge section. However, at this magnification it is impossible to resolve whether individual fibres remain load bearing as the crack crosses the fibre bundles. On the local scale, cracks appear to be more prevalent within the transverse fibre bundles rather than regions of continuous matrix.

[ Zoom ]
Figure 3. Progressive damage accumulation detected on a single sub-surface plane at 25%, 50%, 75% and 100% of applied Pmax.
Figure 3. Progressive damage accumulation detected on a single sub-surface plane at 25%, 50%, 75% and 100% of applied Pmax.

Detailed manual inspections through the XCT image slices, viewed at numerous angles to the specimen orientation and compared back to the unloaded reference image, helped conclude that the initiation of cracks occurred sometime between the first and second load step. This is illustrated by XCT images taken at three different depths (designated planes n0, n1 and n2) relative to the front face in Figure 4, where the unloaded reference image is illustrated in image a), the same plane at 50% Pmax in b) and finally the same plane again under Pmax in image c). The density of cracking appears greatest on plane n2.

[ Zoom ]
Figure 4. XCT images from three sub-surface planes; a) prior to loading
Figure 4. XCT images from three sub-surface planes; a) prior to loading
[ Zoom ]
Figure 4. XCT images from three sub-surface planes; b) under 50% Pmax,
Figure 4. XCT images from three sub-surface planes; b) under 50% Pmax
[ Zoom ]
Figure 4. XCT images from three sub-surface planes; c) under Pmax
Figure 4. XCT images from three sub-surface planes; c) under Pmax

The beneficial aspects of in-situ XCT inspection for the detection of damage are exemplified by Figure 5. Under the application of Pmax a network of cracks is obvious, evenly distributed throughout the gauge length on this specific sub-surface plane. Viewing the same plane after unloading the specimen back to zero, however, indicates that many of these cracks have effectively “closed” now making their detection impossible. Closure occurs across cracks formed in both fibre and matrix dominated regions.

[ Zoom ]
Figure 5. Crack network imaged under Pmax (left) and after unloading (right).
Figure 5. Crack network imaged under Pmax (left) and after unloading (right)

4. Discussion

The employment of a commercial mechanical test rig designed to allow XCT inspection of small scale specimens under in-situ loading was applied to a CMC specimen with notable success. Monotonic tensile stresses were applied in a stepped fashion, with pre-determined hold periods imposed to allow CT data to be taken whilst the specimen was held under a fixed displacement. The initiation of cracks was detected beyond a threshold load condition, with individual cracks then monitored throughout the remaining sequence of loading.

From the perspective of mechanical understanding, the current experiment has met key requirements often expected in support of engineering component design. The specimen gauge volume sampled a reasonable number of fibre bundles and plies in all three dimensions and stress was applied through mode 1 fracture where tensile stress is applied normal to the plane of the crack. Such demands are not always satisfied by sophisticated imaging experiments.

In terms of XCT data acquisition, imaging was affected due to the attenuation of the X-rays penetrating the glassy carbon tube employed by the Deben system. In addition, given the relatively limited space between the opposing metal grips to either end of the CMC specimen, a degree of X-ray reflection was also observed towards the extremities of the field of view. However, the resolution achieved has allowed for the detection of fine scaled cracking throughout the critically stressed gauge section of the specimen.

The in-situ, loaded nature of the present experiment is clearly an important benefit in detecting cracking. Post mortem XCT inspections of tensile, creep or fatigue loaded CMC specimens of varying composition, previously conducted amongst the collaborating partners, have hitherto failed to identify such damage so widely spread away from the ultimate plane of fracture(8). To date, transverse cracking has been thought to be restricted to regions immediately adjacent to the critical crack, with a typical example reproduced in Figure 6 for example. However, the presence of a broader network of cracks is actually consistent with evidence of widely distributed strain “hotspots” as detected up to the onset of failure in two dimensional imaging of the front face of CMC specimens using digital image correlation (DIC) and the spatial distribution of acoustic emission events detected during the same previous tests(9). The three dimensional capability and greater resolution of XCT offers obvious advantages for visualising cracks when compared to DIC techniques and acoustic emission provides no direct physical measurement of crack size.

[ Zoom ]
Figure 6. Subsidiary transverse cracks immediately adjacent to failure zone
Figure 6. Subsidiary transverse cracks immediately adjacent to failure zone

Imaging of the present specimen after returning back to zero load demonstrated that any pre-existing cracks are subject to closure, making some of them undetectable within the resolution of the current XCT system. This was particularly evident where a crack was located within a confined region of the matrix phase. However, evidence of complete closure was also noted where a crack was traversing 0˚ longitudinal fibre bundles. This may suggest that the extent of fibre/matrix de-cohesion along the length of any bridging fibres was limited, allowing the opposing faces of the crack flanks to fully mate during the unloading half-cycle.

The spatial resolution possible with the current 225kV X-ray system could not detect the form of individual fibres.  The system could not be used to characterise intimate details of crack-fibre interactions, certainly not to the degree achieved through typical destructive sectioning and scanning electron microscopy techniques, illustrated for a SiC-alumina composite in Figure 7. However, future experiments are planned where the Deben loading rig will be interfaced to a Zeiss XRADIA system capable of nano-scale resolution. Naturally, a trade-off for any improved resolution will be expected, with the period required to sample the same critically stressed gauge volume being extended.

[ Zoom ]
Figure 7. Typical high magnification imaging of crack/fibre interactions in a CMC achieved via destructive sectioning and scanning electron microscopy
Figure 7. Typical high magnification imaging of crack/fibre interactions in a CMC achieved via destructive sectioning and scanning electron microscopy

Careful inspection of the progressive XCT imaging beyond the second load step allowed for the extension of numerous but individual cracks to be monitored. Figure 8 illustrates the damage evolution for a selected sub-surface plane. On this specific sub-surface plane two significant curved cracks can be seen towards the top of the specimen initiated from the left hand edge. These have clearly been influenced by the non-linear stress field imposed within the vicinity of the gauge-shoulder radius, demonstrating an arched crack path. From a mechanical testing perspective this indicates a change to the specimen geometry may be considered for future experiments, incorporating a larger blend radius with lower associated stress concentration effect. At the same time an increase in gauge length may also be possible within the axial stroke capacity of the test machine. But in terms of damage, one of these arched cracks illustrates a notable effect. At Pmax (load step 4) the crack intercepts a significant pore within the structure, a remnant of incomplete matrix infiltration during processing. It is interesting that this artefact, despite the relatively large scale, did not act as an initiator of cracking itself.

[ Zoom ]
Figure 8. Damage evolution on a single sub-surface plane (inherent pore circled)
Figure 8. Damage evolution on a single sub-surface plane (inherent pore circled)

Should future in-situ inspections be required at increased stress levels, without down scaling the specimen cross section, then developments to the mechanical test frame would need consideration since the Deben rig employed here was taken close to maximum load capacity. This is one option currently under consideration. In the meantime, repeat experiments are planned similar to that reported here in order to gain confidence in the mechanical response of the load rig and the reproducibility of the damage mechanisms.

5. Conclusions

The present paper reports on our initial trials of in-situ XCT inspections of a nominal CMC material under static tensile loading. Based on the encouraging results a number of interim conclusions are drawn;

  • A parallel sided CMC specimen was tested under tensile stress using a commercially available in-situ loading rig. Informative XCT images were recorded at various stages during loading.
  • The 225kV X-ray system used was capable of detecting cracking. Whilst the resolution of the XCT images was below that required to identify individual reinforcing fibres, interactions between cracking with process artefacts and fibre bundles were visible.
  • The high density of cracks distributed throughout the critically stressed gauge section was a unique observation when compared to previous post mortem inspections of mechanically tested specimens.
  • Crack closure after specimen unloading was observed, explaining why such damage was not previously detected when inspecting specimens under zero stress.
  • Improved resolution should be achieved when interfacing the existing load rig with a nano-focus X-ray system.
  • The design of a bespoke mechanical load system interfaced to a X-ray facility is under consideration to allow characterisation at higher magnitudes of stress without moving to a smaller specimen cross section.


This work was supported through funding from the Welsh Government and the Higher Education Funding Council for Wales as part of the Sêr Cymru I programme. (NRN AEM National Research Network Wales for Advanced Engineering and Materials) [Grant number NRNC04].


  1. A. Kelly, ‘Concise Encyclopedia of Composite Materials’, Elsevier, Dec 2012.
  2. J. Kim and P.K. Liaw, ‘The Nondestructive Evaluation of Advanced Ceramics and Ceramic-Matrix Composites’, JOM, Vol. 50, No. 11, Nov 1998.
  3. K. G. Dassios, E.Z. Kordatos, D.G. Aggelis and T.E. Matikas, ‘Nondestructive Damage Evolution in Ceramic Matrix Composites for Aerospace Applications’, The Scientific Word Journal, Vol.13, June 2013.
  4. E.R. Generazio, NASA Technical Memorandum 105807 ‘Nondestrcutive Evaluation of Ceramic and Metal Matrix Composites for NASA’s HITEMP and Enabling Propulsion Materials Programs’, August 1992.
  5. R.E. Johnston, M.R. Bache, M. Amos, D. Liaptsis, R. Lancaster, R. Lewis, P. Andrews and I. Edmonds, ‘Investigation of non-destructive evaluation methods applied to oxide/oxide fibre reinforced ceramic matrix composite’, in Mechanical Properties and Performance of Engineering Ceramics and Composites VII (eds D. Singh, J. Salem, M. Halbig and S. Mathur), John Wiley & Sons, Inc., Hoboken, NJ, USA. doi: 10.1002/9781118217467.ch6
  6. L. Saucedo-Mora, M. Mostafavi, D. Khoshkhou, C. Reinhard , R. Atwood , S. Zhao, B. Connolly and T. J. Marrow, ‘Observation and simulation of indentation damage in a SiC-SiC fibre ceramic matrix composite’, Finite Elements in Analysis and Design, Vol.110, pp 11-19, 2016
  7. Z. Quiney, M.R. Bache, J.P. Jones, ‘Inspection of SiC/SiC fibre ceramic matrix composite specimens employed for fatigue experiments via laboratory X-ray computed tomography’, 7th International Symposium on NDT in Aerospace, Mo.5,A.4. Bremen, Germany, 16-18 November 2015.
  8. E. Sackett and M. Bache, ‘Damage accumulation under static and fatigue loading in a SiCf/Al2O3 CMC’, in Proc. of 9th European Interregional Conference on Ceramics, Bardonecchia, Italy, September 5-7, 2004.
  9.  C. D. Newton, A. L. Chamberlain, J. P. Jones and M. R. Bache, ‘Inhomogeneous strain distribution in SiCf/SiC coupons under tensile loading’, Proceedings of ASME Turbo Expo 2017: Turbomachinery Technical Conference and Exposition, GT 2017-63045, June 26-30, 2017, Charlotte, NC, USA

For more information please email: